Entropy Based Sampling and Parallel CoT Decoding

Official Repo for Open-Reasoner-Zero

Search-R1: An Efficient, Scalable RL Training Framework for Reasoning & Search Engine Calling interleaved LLM based on veRL

RAGEN leverages reinforcement learning to train LLM reasoning agents in interactive, stochastic environments.

Agent-R1: Training Powerful LLM Agents with End-to-End Reinforcement Learning

ReCall: Learning to Reason with Tool Call for LLMs via Reinforcement Learning

Fine-tuning & Reinforcement Learning for LLMs. 🦥 Train OpenAI gpt-oss, DeepSeek-R1, Qwen3, Gemma 3, TTS 2x faster with 70% less VRAM.

verl: Volcano Engine Reinforcement Learning for LLMs

Fully open reproduction of DeepSeek-R1

Minimal reproduction of DeepSeek R1-Zero

Scalable RL solution for advanced reasoning of language models

A curated reading list of research in Adaptive Computation, Inference-Time Computation & Mixture of Experts (MoE).

awesome grounding: A curated list of research papers in visual grounding

[ECCV 2024] Official implementation of the paper "Semantic-SAM: Segment and Recognize Anything at Any Granularity"

Refine high-quality datasets and visual AI models

git extension for {collaborative, communal, continual} model development

🏆 A ranked gallery of awesome streamlit apps built by the community

Lexical Substitution Framework

OpenAssistant is a chat-based assistant that understands tasks, can interact with third-party systems, and retrieve information dynamically to do so.

RWKV (pronounced RwaKuv) is an RNN with great LLM performance, which can also be directly trained like a GPT transformer (parallelizable). We are at RWKV-7 "Goose". So it's combining the best of RN…

A method to increase the speed and lower the memory footprint of existing vision transformers.

A playbook for systematically maximizing the performance of deep learning models.

GLM-130B: An Open Bilingual Pre-Trained Model (ICLR 2023)

A minimal PyTorch re-implementation of the OpenAI GPT (Generative Pretrained Transformer) training

Pre-trained V+L Data Preparation

CLIP Itself is a Strong Fine-tuner: Achieving 85.7% and 88.0% Top-1 Accuracy with ViT-B and ViT-L on ImageNet

EVA Series: Visual Representation Fantasies from BAAI

PyTorch code for BLIP: Bootstrapping Language-Image Pre-training for Unified Vision-Language Understanding and Generation

[CVPR 2022] Official code for "Unified Contrastive Learning in Image-Text-Label Space"